"New to this" keeps what the labs announced and the models they shipped, and hides the version
bumps, retirement notices and outage reports. Nothing gets reworded either way.
The words the labs use
These turn up in the headlines above. The definitions are ours, the headlines are theirs.
Model
The thing that does the work, like GPT-6 or Claude Opus 5. Labs ship new ones constantly and give them version numbers, so "3.8" is a different model from "3.5", not a patch of it.
Token
How the labs count text for billing. Roughly three quarters of a word in English, so 1,000 words costs you about 1,333 tokens.
Context window
How much text a model can hold in its head at once, counted in tokens. A big window is what lets you paste a whole book in and ask about it.
API
The way a program talks to a model, as opposed to you typing into a chat box. It's billed per token, and it's a separate product from the monthly chat subscription.
Deprecation
A lab announcing that a model will stop working on a given date. If you built something on it, that date is your deadline.
Changelog
The running list of what changed in a piece of software, written by the people who changed it.
Nothing here was an announcement or a new model. Switch back to Everything
to see what the labs did post.
New Features
GPT-6-Astra is now available in the model picker and Amazon Bedrock catalogs. ( #42879 , #42619 )
Experimental worktree support lets you create isolated checkouts for new or forked sessions using --worktree or /worktree , then browse and resume them. ( #42652 , #43069 , #43120 , #43286 )
Answer questions inline while Codex continues working, using suggested choices or custom text without losing your main draft. ( #42891 , #42894 , #42897 )
Windows sessions can now share a background Codex server, with daemon lifecycle commands and managed updates. ( #42405 , #42392 )
Vim editing…
Chris Lehane argues that stronger AI capabilities require stronger safety evidence, shared standards, and durable policy action while the policy window remains open.
What's changed
Added maxEffortLevel setting (top-level or per model under modelSettings ): caps the effort level on every provider, including Bedrock, Vertex and Foundry; users can still pick a lower level
Added --system-prompt-snapshot off to render the system prompt fresh on every request instead of reusing the conversation's recorded prompt (for iterating on prompt text)
Fixed Cowork scheduled tasks in the cloud failing at startup for organizations whose managed settings require sandboxing
Fixed /context and other local command output rendering blank on mobile clients
Fixed shift+enter and…
Added maxEffortLevel setting (top-level or per model under modelSettings): caps the effort level on every provider, including Bedrock, Vertex and Foundry; users can still pick a lower level
Added --system-prompt-snapshot off to render the system prompt fresh on every request instead of reusing the conversation's recorded prompt (for iterating on prompt text)
Fixed Cowork scheduled tasks in the cloud failing at startup for organizations whose managed settings require sandboxing
Fixed /context and other local command output rendering blank on mobile clients
Fixed shift+enter and…
An illustrated graphic set against a vibrant green background featuring American football elements, including a gold trophy, a blue helmet, a silver whistle, a football, a mini scoreboard, and play diagrams, with the icon for AI Mode in Google Search in t
Every line links to the lab that published it. We collect from their blogs, changelogs, status pages and model docs every fifteen minutes, and we don't reword what they wrote. The live feed has the filters and the RSS.